Alright, let's tackle your requirement step by step. You want to generate date-suffixed indexes (like myindex-2024.05.20) automatically when uploading data daily, tie all these indexes to the same ILM policy for easy deletion, and use aliases for seamless writing/querying. Here's how to adjust your existing setup to make this work:

Option 1: Auto-create date-named indexes daily (no rollover needed) #

This approach directly creates a new index for each calendar day, with no dependency on data size. All indexes will be managed by your ILM policy and can be queried through a unified alias.

1. Update your ILM policy (simplify for daily indexes) #

Since we're creating indexes per day, we don't need the rollover action in the hot phase anymore—just keep the delete phase to auto-purge old data:

PUT _ilm/policy/my_policy
{
  "policy": {
    "phases": {
      "hot": {
        "actions": {}
      },
      "delete": {
        "min_age": "180d",
        "actions": {
          "delete": {}
        }
      }
    }
  }
}

2. Create an index template for daily indexes #

This template ensures every new date-suffixed index inherits your settings, ILM policy, and is added to a query alias:

PUT _template/my_daily_index_template
{
  "index_patterns": ["myindex-*"],
  "settings": {
    "number_of_shards": 1,
    "number_of_replicas": 1,
    "index.lifecycle.name": "my_policy",
    "index.lifecycle.origination_date": "{{now/d}}" // Use today's date to calculate ILM age correctly
  },
  "aliases": {
    "myindex-all": {} // Alias to query all historical daily indexes
  }
}

3. Write data to the daily index in Java #

Use Elasticsearch's date math syntax or directly generate the daily index name in your code to auto-create the day's index on first write:

// Option 1: Use Elasticsearch date math (auto-resolves to today's index)
String indexName = "<myindex-{now/d}>";

// Option 2: Generate the date-suffixed index name manually
String formattedDate = LocalDate.now().format(DateTimeFormatter.ofPattern("yyyy.MM.dd"));
String indexName = "myindex-" + formattedDate;

// Then perform your index request as usual
IndexRequest request = new IndexRequest(indexName);
request.id("unique-id-1");
request.source("field1", "value1", "field2", "value2", XContentType.JSON);
client.index(request, RequestOptions.DEFAULT);

Option 2: Use ILM Rollover for daily + size-based index creation #

If you want to keep the flexibility of rolling over indexes when they hit your 25GB limit (even within a single day) while still using date-based naming, use this setup:

1. Update ILM policy to trigger rollover on size OR age #

This ensures a new index is created either when it hits 25GB or at the end of the day:

PUT _ilm/policy/my_policy
{
  "policy": {
    "phases": {
      "hot": {
        "actions": {
          "rollover": {
            "max_size": "25GB",
            "max_age": "1d"
          }
        }
      },
      "delete": {
        "min_age": "180d",
        "actions": {
          "delete": {}
        }
      }
    }
  }
}

2. Create an initial rollover index #

Start with a date-prefixed index, link it to your write alias, and configure rollover settings:

PUT myindex-2024.05.20-000001
{
  "aliases": {
    "myindex-write": {
      "is_write_index": true
    }
  },
  "settings": {
    "number_of_shards": 1,
    "number_of_replicas": 1,
    "index.lifecycle.name": "my_policy",
    "index.lifecycle.rollover_alias": "myindex-write",
    "index.lifecycle.origination_date": "{{now/d}}"
  }
}

3. Create a template for rollover indexes #

Ensure all future rolled-over indexes inherit your settings and join the query alias:

PUT _template/my_rollover_template
{
  "index_patterns": ["myindex-*-*"],
  "settings": {
    "number_of_shards": 1,
    "number_of_replicas": 1,
    "index.lifecycle.name": "my_policy",
    "index.lifecycle.rollover_alias": "myindex-write"
  },
  "aliases": {
    "myindex-all": {} // Unified alias for querying all rolled-over indexes
  }
}

4. Write data via the write alias #

In your Java code, just write to the myindex-write alias—ILM will automatically handle switching to the new index when rollover conditions are met:

IndexRequest request = new IndexRequest("myindex-write");
request.id("unique-id-1");
request.source("field1", "value1", XContentType.JSON);
client.index(request, RequestOptions.DEFAULT);

Managing Deletions #

• Manual delete a single daily index: Run DELETE myindex-2024.05.20
• Auto-delete old indexes: Your ILM policy will automatically delete indexes older than 180 days
• Delete all indexes matching the pattern: DELETE myindex-* (use this cautiously!)

内容的提问来源于stack exchange，提问作者GKr297